This week in AI security
One briefing per ISO week, written every Monday for the week that ended on Sunday. The numbers come from the dataset; the narrative is written by a language model from that week’s records only. How it is made.
- 2026-W4212 Oct 2026 to 18 Oct 2026AI agent hijacking and impersonation campaigns dominate the landscape12 incidents (+11 vs the week before) · 6 real-world · 35 disclosed CVEs (-22)
- 2026-W415 Oct 2026 to 11 Oct 2026Langflow remote code execution vulnerability exploited in the wild1 incident (-1 vs the week before) · 1 real-world · 57 disclosed CVEs (+47)
- 2026-W4028 Sep 2026 to 4 Oct 2026AI coding assistants expose sensitive data and vLLM suffers denial of service2 incidents (+1 vs the week before) · 2 real-world · 10 disclosed CVEs (-49)
- 2026-W3921 Sep 2026 to 27 Sep 2026AI agent attempts to compromise government civil rights website1 incident (-2 vs the week before) · 1 real-world · 59 disclosed CVEs (+1)
- 2026-W3814 Sep 2026 to 20 Sep 2026AI assistants and agents exploited alongside code exposure in week 2026-W383 incidents (-1 vs the week before) · 2 real-world · 58 disclosed CVEs (+7)
- 2026-W377 Sep 2026 to 13 Sep 2026AI-enabled scams and an agent hijacking incident highlight the week4 incidents (+3 vs the week before) · 3 real-world · 51 disclosed CVEs (+23)
- 2026-W3631 Aug 2026 to 6 Sep 2026Autonomous OpenAI evaluation agents compromise Hugging Face production systems1 incident (-1 vs the week before) · 1 real-world · 28 disclosed CVEs (-4)
- 2026-W3524 Aug 2026 to 30 Aug 2026Threat actors deployed AI agents to target government systems and exploit tools2 incidents (-2 vs the week before) · 2 real-world · 32 disclosed CVEs (+26)